The IMF and the Global Financial Crisis

Regulating  Capital  Flows  in  Emerging  Markets:   the  IMF  and  the  Global  Financial  Crisis   Abstract   In  the  wake  of  the  financial  crisis  the  International  Monetary  Fund  (IMF)  began  to   publicly  express  support  for  what  have  traditionally  been  referred  to  as  ‘capital   controls’.    In  addition  to  public  statements,  the  IMF  underwent  a  systematic  re-­‐ evaluation  of  Fund  policy  on  the  matter,  and  published  an  official  view  on  the   economics  of  capital  flows.    In  this  view  the  IMF  concluded  that  capital  account   liberalization  is  not  always  the  most  optimal  policy  and  that  there  are  situations   where  capital  controls—rebranded  as  ‘capital  flow  management  measures’—are   appropriate.    This  paper  empirically  examines  the  extent  to  which  the  change  in  IMF   discourse  on  these  matters  has  resulted  in  significant  changes  in  IMF  policy  advice.     To  answer  this  question  we  create  a  database  of  IMF  Article  IV  reports  and  examine   whether  the  financial  crisis  had  an  independent  impact  on  IMF  support  for  capital   controls.    We  find  that  the  IMF’s  level  of  support  for  capital  controls  has  increased  as   a  result  of  the  crisis  and  as  the  vulnerabilities  associated  with  capital  flows   accentuate.   1   1. Introductory  Framework  and  Literature  Review Sometimes  financial  crises  make  policy-­‐makers  stop  and  rethink  whether  they  know   what  they  think  they  know  about  how  economies  work  and  what  the  proper   economic  policy  responses  should  be  to  prevent  and  mitigate  such  crises.    Was  this   time  different?    It  has  been  well  established  that  the  International  Monetary  Fund   (IMF),  though  no  longer  wholly  opposed,  was  generally  skeptical  for  the  regulation   of  cross-­‐border  financial  flows  from  the  1980s  to  the  run  up  to  the  global  financial   crisis  (Abdelal,  2007;  Chweiroth,  2009;  Moschella,  2010).    In  the  wake  of  the  crisis   the  IMF  surprised  many  observers  by  openly  embracing  capital  controls  to  both   prevent  and  mitigate  financial  crises.  The  IMF  supported  the  use  of  capital  controls   on  inflows  in  a  number  of  countries  such  as  Brazil  and  South  Korea.    Most  surprising   to  many  was  the  IMF’s  strong  support  of  the  use  of  capital  controls  on  outflows  in   Iceland  as  part  of  that  country’s  post  crisis  stand-­‐by-­‐agreement  (Sigurgeirsdottir   and  Wade,  2014).      A  burgeoning  literature  has  explained  the  role  the  crisis  played   in  the  shift  of  discourse  at  the  IMF  on  this  matter  (Grabel,  2011;  Chweiroth,  2013;   Gallagher,  2014).    It  is  clear  that  the  crisis  played  an  independent  part  in  at  least   accelerating  an  incremental  level  of  ideational  change  at  the  fund  on  this  issue,   though  the  seeds  of  change  were  planted  after  the  wave  of  crises  that  ended  the   century.    This  paper  takes  such  analyses  one  step  further  by  analyzing  the  extent  to   which  such  changes  in  discourse  related  to  the  crisis  were  also  associated  with   changes  in  official  IMF  advice  on  managing  capital  flows.       A  strand  of  theory  in  the  international  political  economy  literature  postulates  that   during  episodes  of  financial  crises  that  firmly  held  ideas  can  be  challenged  by  a  rival   set  ideas  about  how  economies  work  and  should  be  managed.    Under  the   uncertainty  that  is  rife  in  such  episodes,  certain  key  agents  can  be  open  to   alternative  ideas  that  help  manage  such  uncertainty.    The  conduits  for  such  change   can  be  ‘norm  entrepreneurs’’  that  cultivate  ‘pervasive  struggles’  to  legitimize   previously  unaccepted  views  (Blyth  2002;  Seabrooke  2007;  Schmidt  2008;   Widmeier  et  al  2007).      In  the  global  uncertainty  following  the  global  financial  crisis   a  significant  amount  of  research  demonstrates  that  the  IMF  changed  the  way  it   talked  about  global  capital  flows  and  their  benefits  and  risks.       In  the  1990s  the  IMF  underwent  a  paradigm  shift  and  began  to  see  capital  account   liberalization  as  an  optimal  policy  for  all  countries,  and  thus  saw  capital  controls  as   an   unadvisable   policy.     Indeed,   in   the   1990s   the   IMF   went   so   far   as   to   introduce   a   formal  change  to  its  Articles  of  Agreement  that  would  have  mandated  open  capital   accounts   for   its   membership.     As   a   result   of   the   financial   crises   of   the   1990s,   and   actions   by   the   United   States   Congress,   that   proposal   did   not   come   to   fruition.   Subsequently,   the   IMF   became   more   tolerant   of   the   gradual   liberalization   of   the   capital   account   and   of   temporary,   price-­‐based   capital   controls   as   a   last   resort   for   emerging   market   and   developing   countries   (Independent   Evaluation   Office,   2005;   Abdelal,  2007;  Chweiroth,  2009;  Moschella,  2010).   2   A   significant   shift   in   mainstream   economic   thinking   regarding   the   regulation   of   capital  flows  occurred  around  the  time  of  the  crisis  as  well.    Mainstream  economic   thought   generally   saw   capital   account   liberalization   as   an   optimal   policy   in   the   long   run   for   all   countries   and   saw   the   regulation   of   capital   flows   as   inherently   distortionary   from   that   optimum.     Certain   strands   of   economics   from   the   Keynes,   Minsky,  and  Lewis  traditions  had  long  seen  the  regulation  of  capital  as  necessary  for   maintaining  monetary  policy  autonomy,  preventing  financial  fragility,  and  as  levers   for   structural   transformation.       These   perspectives   had   fallen   out   of   the   mainstream   by  the  1980s  (Gallagher,  2014).   Around  the  time  of  the  global  financial  crisis  a  consensus  among  mainstream  began   to  emerge  on  both  the  theory  and  the  econometric  evidence  related  to  capital   account  liberalization  and  the  regulation  of  capital  flows.    A  number  of  theorists   began  to  question  the  extent  to  which  capital  account  liberalization  is  optimal,   especially  in  the  presence  of  information  externalities.    According  to  this  research,   externalities  are  generated  by  capital  flows  because  individual  investors  and   borrowers  do  not  know  (or  ignore)  what  the  effects  of  their  financial  decisions  will   be  on  the  level  of  financial  stability  in  a  particular  nation.  This  is  a  classic  market   failure  argument  and  calls  for  what  is  referred  to  as  a  Pigouvian  tax  that  will  correct   for  the  market  failure  and  make  markets  work  more  efficiently.    These  theoretical   breakthroughs  were  further  substantiated  given  that  the  vast  majority  of   econometric  analyses  of  capital  account  liberalization  find  no  rigorous  link  between   capital  account  liberalization  and  growth  in  emerging  market  and  developing   countries,  and  that  liberalization  is  often  linked  to  banking  crises  (Jeanne  et  al,   2012).    Finally,  meta-­‐reviews  of  the  literature  on  the  effectiveness  as  capital  controls   found  that  capital  controls  consistently  had  the  desired  effects  of  their  policy-­‐ makers  (Magud  et  al,  2011;  Jeanne  et  al,  2012).    An  authoritative  review  of  the   literature  on  these  matters  concluded  the  following:   “The  international  community  should  not  seek  to  promote  totally  free  trade   in  assets—even  over  the  long  run—because  (as  we  show  in  this  book)  free   capital  mobility  seems  to  have  little  benefit  in  terms  of  long-­‐run  growth  and   because  there  is  a  good  case  to  be  made  for  prudential  and  other  non-­‐ distortive  capital  controls.”  (Jeanne  et  al,  2012,  5).   The  IMF  took  an  even  larger  step  in  accepting  gradual  capital  account  liberalization   and   the   use   of   capital   controls   in   the   wake   of   the   global   financial   crisis   of   2008.     First,   it   is   important   to   note   that   the   crisis   was   associated   with   significant   surges   and   sudden   stops   in   cross-­‐border   capital   flows   as   Figure   1   shows,   there   was   a   sudden   stop   in   capital   flows   to   emerging   market   and   developing   countries   as   a   result  of  the  crisis—with  investors  flocking  to  the  ‘safety’  of  industrialized  markets.       3   Figure  1  All  Emerging  Market:  Gross  Capital  Inflows  1998-­‐2012   30   25   15   Median   10   25th  percentile   5   75th  percentile   0   -­‐5   -­‐10   1998q1   1999q1   2000q1   2001q1   2002q1   2003q1   2004q1   2005q1   2006q1   2007q1   2008q1   2009q1   2010q1   2011q1   2012q1   %  of  GDP 20   Source:  IMF,  International  Financial  Statistics,  2014   However,  as  nations  such  as  the  United  States  engaged  in  expansionary  monetary   policy,  investors  again  began  to  surge  into  emerging  markets.    It  is  under  this   turbulent  period  that  then  managing  director  Dominique  Strauss  Kahn  ignited  a   sense  of  new  thinking  within  the  Fund  in  hopes  that  it  would  revive  interest  in  the   IMF,  given  that  global  regard  for  the  institution  had  waned  significantly.    Norm   entrepreneurs  within  the  research  department  seized  that  moment  and  published   articles  that  found  that  those  countries  that  deployed  capital  controls  going  into  the   crisis  were  among  the  least  hard  hit  (Ostry,  et  al,  2010).  These  findings  were   supported  and  promoted  by  the  managing  director  and  led  to  an  eventual  official  re-­‐ evaluation  of  the  IMF  position  on  capital  account  liberalization  and  capital  controls.     This  re-­‐evaluation  was  hotly  contested  within  the  board  of  the  IMF,  with  the  BRICS   countries  leading  an  efforts  to  grant  the  most  policy  space  possible  for  emerging   markets  to  regulate  capital  flows  (Chwieroth,  2013;  Gallagher,  2014).  In  December   2012,  IMF  adopted  a  ‘New  Institutional  View’  on  capital  flow  management  (IMF,   2012).    In  the  new  view,  the  IMF  now  recognizes  that  capital  flows  carry  risks  and   that  the  liberalization  of  capital  flows  before  nations  reach  a  certain  threshold  of   financial  and  institutional  development  can  accentuate  those  risks.  The  IMF  also   now  acknowledges  that  under  certain  circumstances,  cross-­‐border  capital  flows   should  be  regulated  to  avoid  the  worst  effects  of  capital  flow  surges  and  sudden   stops.  These  tenets  were  incorporated  into  a  Staff  Guidance  note  in  2013  and  since   that  time  are  intended  to  guide  official  IMF  policy  advice  on  the  matter  (Grabel   2011;  Chwieroth,  2013;  Gallagher,  2014).   While  there  is  an  emerging  literature  on  the  extent  to  which  the  IMF  has  changed  its   policy  and  discourse  with  respect  to  managing  capital  flows,  there  is  yet  systematic   research  that  quantitatively  examines  the  extent  to  which  the  IMF  has  actually   changed  its  policy  advice.    There  is  a  significant  literature  that  attempts  to  quantify   4   the  extent  to  which  the  IMF  has  changed  its  behavior  in  other  issues.    Vreeland   (2003),  Pop  Ecles  (2008),  Thacker  (2009),  Bird  and  Rowlands  (2009),  and   Presbitero  and  Zazzaro  (2012)  have  all  examined  the  quantitative  determinants  of   IMF  lending  programs  in  different  settings.       There  is  also  one  article  that  empirically  examines  the  relationship  between  the  IMF   and  policy  on  capital  flows.    Joyce  and  Noy  (2008)  empirically  examine  the  extent  to   which  IMF  country  programs  in  the  1980s  and  1990s  were  associated  with  policies   to  liberalize  the  capital  account.    The  authors  do  indeed  find  evidence  that  IMF   programs  were  correlated  with  capital  account  liberalization.    Roy  and  Ramos   (2012)  examine  Article  IV  reports  to  identify  whether  the  IMF  has  changed  its  policy   advice  after  the  crisis.    For  their  paper  they  read  26  reports  in  2010  and  did  not  see   much  of  a  change  in  IMF  behavior.    In  the  spirit  of  Joyce  and  Noy,    our  paper  is  the   first  to  our  knowledge  that  econometrically  examines  the  extent  to  which  the  IMF   has  demonstrably  changed  its  advisory  behavior  as  manifest  in  official  Article  IV   reports  both  before  and  after  the  global  financial  crisis.         2. Data  and  Methodology The  specific  research  questions  for  this  paper  are:  To  what  extent  has  the  IMF   changed  the  way  it  views  capital  flows,  and  to  what  extent  has  the  IMF  increased  its   level  of  support  for  capital  controls  in  the  wake  of  the  financial  crisis?  This  section  of   the  paper  describes  the  mechanics  of  the  database  that  was  created  to  answer  our   research  question  and  outlines  the  econometric  model  and  methodology  for  the   research  as  a  whole.   2.1  Database   Our  study  is  based  on  a  unique  dataset  created  from  IMF  annual  Article  IV  reports   since   1998.   We   built   a   database   focusing   on   capital   flow   management   and   related   policies   for   31   emerging   markets   covering   Asia,   Latin   America   and   Caribbean,   Europe   and   Africa.   This   database   includes   coding   IMF   Article   IV   Consultation   Reports   and   Public   Information   Notice   well   as   collecting   country-­‐specific   macroeconomic  data  from  the  World  Bank  World  Development  Indicators  database   (IMF,  2013;  World  Bank,  2014).     Our  coding  method  is  derived  from  a  2005  IMF  Evaluation  Report  prepared  by  the   Independent   Evaluation   Office   titled   The   IMF's   Approach   to   Capital   Account   Liberalization  (IMF,  2005).  In  that  report,  through  in  a  more  qualitative  manner  and   with   a   smaller   set   of   countries,   the   IMF’s   Independent   Evaluation   Office   (IEO)   assessed   the   level   of   IMF   support   for   capital   account   liberalization   and   capital   controls  in  the  wake  of  the  financial  crises  in  the  1990s.    Juxtaposed  with  the  new   data   we   derive   from   IMF   Article   IV   reports,   we   include   macroeconomic   data   such   as   current   account   balance,   domestic   credit   of   the   banking   sector   and   external   debt   5   payments.   Overall,   our   database   includes   528   observations   of   33   countries   in   16   years  (1998-­‐2013).   2.1.1  Coding  Criteria   Figure   1   shows   the   coding   process   and   the   coding   manual   is   in   the   online   appendix.   For   each   Article   IV   report,   we   first   code   the   IMF   initial   diagnosis   by   examining   whether  the  IMF  deems  capital  flows  as  an  area  of  concern  for  a  country  undergoing   an  Article  IV  consultation.  Key  words  were  searched  and  read  for  were  those  such  as   ‘capital   flow   measures,’   ‘capital   controls,’   ‘financial   stability,’   ‘surge,’   ‘sudden   stop,’   ‘unremunerated   reserve   requirement,’   ‘capital   account   deficit,’   ‘inflows/outflows,’   ‘exchange  rate  risk,’  ‘debts  denominated  in  foreign  currency,’  and  others.   Keywords   of   a   capital   flow   issues   include   ‘external   shock,’   ‘external   instability,’   ‘adverse  shocks  related  to  global  stress,’  ‘adverse  spillovers  arising  from  the  global   turmoil,’   ‘contagion,’   ‘foreign   exchange   pressures,’   ‘rising   external   imbalance’,   ‘external  financial  environment,’  ‘balance  of  payment  pressure.’   Secondly,  we  code  the  IMF’s  policy  recommendations  to  remedy  concerns  related  to   capital  flows.    We  code  each  policy  recommendation  separately,  corresponding  with   the  measures  coded  by  the  IEO  in  their  2005:  Tighten  Fiscal  Policy,  Exchange  Rate   Flexibility,   Sterilization/Intervention   in   the   currency   market,   Trade   Liberalization,   Tighten  Prudential  Regulation,  Capital  Flows  Management/Capital  Controls.       Finally,   if   the   IMF   has   a   policy   recommendation   with   respect   to   Capital   Flow   Management   measures   or   capital   controls,   we   code   the   IMF’s   level   of   support   for   such   measures.   A   common   response   to   managing   capital   flows   is   to   tighten   fiscal   policy.   Any   recommendation   to   engage   in   fiscal   tightening   is   in   response   to   macroeconomic   factors.   Key   words   in   tightening   fiscal   policy   include   overheating   economy,  risk,  ease  burden  on  monetary  policy,  cushion  to  weather  shock,  response   to  a  possible  surge  in  capital  inflows,  inflation  resulting  from  an  external  situation.       Exchange  rate  flexibility  is  also  advocated  by  the  IMF  and  others  as  tool  to  temper   swings   in   capital   flows.     A   flexible   exchange   rate   can   be   a   shock   absorber   in   the   event   of   capital   inflow  surge.   For   sterilization   /intervention   in   the   foreign   exchange   markets,   we   read   for   endorsements   of   a   reserve   build-­‐up,   higher   reserve   levels   would   help   guard   against   capital   account   shocks,   intervention   in   FX   markets   to   smooth  volatility  and  enhance  liquidity,  reserve  accumulation,  purchases  of  foreign   exchange.     Tighten  macro  prudential  policy  key  words  are  capital  requirements  for  banks,  bank   soundness,   systemic   risk,   Basel   III,   improve   supervision   and   regulation,   guard   against   risk,   measures   to   prevent   liquidity   crisis   in   banking   system,   raising   capital   adequacy   ratio,   reinforcing   financial   soundness,   promoting   financial   sector   deepening.     6   We  also  code  IMF  recommendations  with  respect  to  CFMs.  The  key  words  includes:   capital  controls2,  capital  flow  measures,  CFM,  unremunerated  reserve  requirement,   impact  of  capital  controls/  ineffectiveness,  capital  account  regulation  of  a  prudential   nature,  and  so  forth.     There  are  numerous  kinds  of  CFMs  or  capital  controls,  such  as  taxes  on  the  inflow  or   outflow   of   capital,   quantitative   measures   on   the   repatriation   of   portfolio   investments,   exit   levies,;   prohibition   of   foreign   purchase   or   holding   of   domestic   assets;   requirements   to   obtain   administrative   permission   for   a   foreign   bond   issue;   minimum  maturity  period  for  foreign  bond  issues;;  taxes  on  purchases  of  domestic   assets   by   foreigners   or   on   investment   income   earned   by   foreigners;   reserve   requirements  on  deposits  held  by  foreigners  and  others  (see  Gallagher,  2014).   Level   of   support   for   capital   controls   are   divided   into   four   groups:   not   supportive   (phase   out   controls,   controls   are   ineffective,   drawbacks,   elimination   of   controls   as   a   positive   step,   negative   effects   of   capital   controls),   not   mentioned,   partially   supportive   (management   of   temporary   surge,   could   be   an   option,   part   of   a   transitory   response)   and   fully   supportive.     We   code   these   as   -­‐1,   0,   1,   and   2,   respectively.    Figure  1  outlines  our  coding  methodology  in  greater  detail.   2  If  controls  are  mentioned  in  the  context  of  FDI,  we   still  code  it  as  a  mention  and  add  comments  in  separate  column,   specifying  the  type  of  capital  flow  (inflows  or  outflows)  and  the  type  of  capital  control.   7   Figure  2  Coding  Process   Coding  Manual  Creation:  develop  codebook  based  on  initial  reading  of  IEO  evalution  report  and  IMF  institutional   views.   Loop  until   pass  the   test   Random  Sample  Coding:  given  the  same  coding  sample  from  database  ,  coders  code  independently.   Random   Sample     Coding   Bias  Control  Test:  inter  coder  statistics  are  calculated  on  the  coding  sample.  Compare  results  and  test  reliability.   Codebook  ModiOication:  discuss  and  modify  problematic  codes.  Modimied  code  manual  is  given  back  to  coders.   Entire     Dataset   Coding   Entire  set:  team  of  coders  is  assigned  countries  in  different  categories.   Reliability  check  and  Oinal  codebook  revision:  coders  discuss  issues  case  by  case  and  point  out  continuing   coding  discrepancies.  Coding  manual  is  modimied  to  account  for  discrepancies.   Final  Coding:  coders  code  independently  and  cover  the  full  set  of  article  IVs.   Reconciliation  and  Merge:  Coders  discuss  discrepancies  and  make  correponding  modimications  to  create  a  minal   dataset   8   2.1.2  Tests  on  Bias  Control   To   control   the   coding   bias   among   different   coders,   we   apply   the   kappa   and   concordance   methods.   These   methods   are   widely   used   in   survey   and   coding   literature,  especially  medical  science.  The  estimators  are  an  inter-­‐coder  agreement   measure   describing   the   reliability   of   the   coding.   What   all   inter-­‐coder   agreement   measures  do  is  adjust  for  this  level  of  chance  in  some  way.   The  standards  for  what  suffices  in  terms  of  an  acceptable  'threshold'  (e.g.  60%  inter-­‐ coder   agreement)   are   well   defined   based   on   specific   cases   and   not   fixed.   The   higher   the  agreement  estimators  are,  the  more  consistent  the  coding  is.  They  give  feedback   on  whether  and  how  to  adjust  the  coding.  In  addition,  this  helps  the  team  of  coders   to  be  on  the  same  page  and  make  precise  quantitative  coding.   The   process   of   coding   at   its   initial   stage   is   an   iterative   process.   After   adapting   the   codebook   and   communicating   together   as   a   group   to   perfect   the   procedure   of   coding,  we  reach  an  inter-­‐coder  reliability  scores  of  88.7%.     To   measure   the   inter   coder   variation   so   that   two   or   more   independent   coders   are   evaluation  the  same  thing,  we  use  kappa  statistic  as  a  numerical  rating  of  the  degree   to  which  the  coding  is  consistent.  First,  five  coders  are  coding  independently  on  the   same   country-­‐-­‐-­‐   Columbia   1998-­‐2012   Article   IVs.   Each   coder   codes   15   observations   on   policy   indicators   and   capital   control   measures.   Second,   coders   discuss   the   differences   and   compare   each   coding   results   with   the   corresponding   sources.   Disagreements   are   settled   on   specific   issues   and   the   team   comes   up   with   the   first   draft  of  coding  manual.  Then,  coders  are  re-­‐coding  again  based  on  the  manual.  After   several   rounds   of   discussion   and   review,   the   coding   results   are   pooled   together   and   kappa  statistics  are  calculated.   The   calculation   is   based   on   the   difference   between   how   much   agreement   is   actually   present   compared   to   how   much   agreement   would   be   expected   to   be   present   by   chance   alone.   Kappa   is   a   measure   of   the   difference   between   the   observed   and   expected  agreement,  standardized  to  lie  on  a  -­‐1  to  1  scale.                       9   Table  1:  Inter  Coder  Variation  on  Level  of  Supports  in  Capital  Controls   Outcome   Kappa  Statistic   Z   Prob>Z   0   0.5076   5.31   0.0000   1   0.3647   3.45   0.0003   2   0.4663   6.08   0.0000   3   0.0119   0.16   0.5617   Combined   0.4280   6.58   0.0000     Table  1  indicates  the  average  outcomes  obtained  from  the  pool  of  five  coders.  Based   on  the  general  criteria  as  shown  in  Graph  1,  the  kappa  statistic  scores  for  0  and  2  are   moderate,   1   is   fair   while   3   is   poor.   The   kappa-­‐statistic   measure   of   agreement   is   scaled   to   be   0   when   the   amount   of   agreement   is   what   would   be   expected   to   be   observed  by  chance  and  1  when  there  is  perfect  agreement.     Figure  3  Interpretation  of  Kappa   Agreement              Kappa:   Poor                      0.0   Slight                  .20   Fair                        .40   Moderate  .60   Substantial        . 80   Almost   Perfect  1.0       The   estimated   kappa   itself   could   be   due   to   chance.   The   P   value   and   z   statistics   of   kappa  reports  the  variance  of  kappa.  We  eliminate  the  case  of  estimated  kappa  due   to  chance  with  statistically  significant  kappa  above  0.     To  evaluate  the  quality  of  coders’  coding,  we  also  calculate  the  correlation  and  apply   the   concord   commend   in   STATA   for   any   two   coders.   Concordance   correlation   coefficient   (Lin,   1989,   2000)   reflects   a   statistic   comparison   between   different   coders.   This   helps   to   find   the   difference   between   two   coders.   Here   we   report   the   correlation  coefficient  of  5  coders  on  coding  Columbia  Article  IVs  in  Table  2.  We  find   coder1   and   coder   3   has   the   highest   correlation   while   there   are   disagreements   among  coder  2  and  coder  4.  Coder  2’s  coding  has  the  relative  poor  quality  and  the   group  discussion  solves  the  disagreement  term  by  term.             10   Table  2:  Correlation  between  raters  on  all  coding  observations   Correlation   Coder0   Coder1   Coder2   Coder3   Coder4   Coder0   1.0000           Coder  1   0.6035   1.0000         Coder  2   0.3671   0.1364   1.0000       Coder  3   0.7493   0.8870   0.3155   1.0000     Coder  4   0.6187   0.5581   0.1527   0.6954   1.0000     After   recoding   the   disagreement   term   together,   we   calculate   the   concordance   correlation   between   coders   again.   The   reason   to   introduce   Lin’s   concordance   method  is  that  it  is  robust  on  as  few  as  10  pairs  of  data,  which  is  more  suitable  in   our  coding  situation.  The  results  are  summarized  in  Table  3.  We  reach  an  agreement   with   concordance   correlation   coefficient   (>0.8:   substantial;   >0.6:   moderate).   The   average  of  differences  between  coders  is  controlled  under  0.2.   Tests   on   bias   control   are   helpful   in   amendments   and   give   suggestions   on   revising.   It’s  important  to  detect  the  difference  before  generating  to  other  countries.  On  the   other   hand,   case   study   also   plays   an   important   role   in   the   coding   process.   It   provides  efficient  coding  process.     In  sum,  the  final  coding  manual  passes  the  moderate  level  of  inter  coder  agreement   statistically.  Our  coding  results  among  all  countries  across  time  are  consistent  and   the   team   has   the   unified   criteria   on   quantifying   the   policy   factors.   The   level   of   supports  on  capital  controls  are  coded  neutrally  according  the  reports  sentence  by   sentence.  The  evaluation  from  5  coders  is  on  the  same  level  statistically,  taking  into   consideration  the  variation  by  chance.                     11   Table  3:  Concordance  Correlation  Coefficient   Comparison   Pairs   Concordance   Coefficient   Correlation   Difference  Statistics   Coder  #   Coder  #   rho_c   SE(rho_c)   95%  CI   Average   Correlation     (asymptotic)   btw   difference   &   mean   rater1   rater0   0.622   0.156   [0.316,   0.929]   -­‐0.059   -­‐0.139   rater1   rater2   0.496   0.240   [0.276,   0.667]   -­‐0.176   -­‐0.093   rater1   rater3   0.872   0.065   [0.744,   1.000]   0.133   0.019   rater1   rater4   0.657   0.167   [0.229,   0.885]   -­‐0.118   -­‐0.318     After  several  rounds  of  kappa  and  concordance  tests,  we  improve  the  coding  manual   and  come  up  with  a  consistent  result.  At  the  end  of  the  coding,  we  go  through  all  the   coding   results   to   check   and   correct   for   consistency,   taking   into   account   the   structural  relationship  between  each  element  of  every  observations.  We  also  fill  0  to   the  blank  cells  to  complete  the  data  set  based  on  the  related  report  reference.   2.2  Econometric  Methodology   Given   the   dataset   coded   from   Article   IV   Reports,   we   apply   OLS   regression   in   a   reduced-­‐form   econometric   model   to   examine   the   extent   to   which   the   IMF   has   changed   its   diagnoses   of   the   role   of   capital   flows   and   whether   the   IMF   has   significantly   changed   its   level   of   support   for   CFMs   as   a   result   of   the   crisis.   Thus,   our   model  is  built  to  analyze  whether  the  IMF’s  view  on  capital  flows  is  affected  by  the   financial  crises  and  the  vulnerability  of  the  economy.  It  also  takes  the  country  fixed   effect  into  consideration  and  runs  the  robust  regression  on  the  panel  dataset  from   2000  to  2013.     The  reasons  for  choosing  the  time  period  after  2000  in  the  regressions  are:  on  one   hand,  IMF  policy  response  to  capital  flows  for  the  1998-­‐2000  periods  is  influenced   by   the   Asian   Financial   Crises;   on   the   other   hand,   during   the   1998-­‐2000   periods,   most  of  the  data  is  based  on  Public  Information  Notice  while  after  2000  most  are  full   Article   IV   reports.   Thus,   to   have   a   more   reliable   data   information   pre   and   post   2008   Financial  Crises,  it  makes  sense  to  choose  the  analysis  period  after  2000.     12   2.2.1  Model  Equations   The   research   question   is   to   examine   the   extent   to   which   the   IMF   policy   advice   on   capital  flows  after  the  financial  crises.  In  other  words,  did  the  2008  global  financial   crises   change   the   IMF   initial   diagnosis   of   capital   flows   issues   for   emerging   markets?   Are   the   Capital   Flow   Management   measures   mentioned   more   frequently   after   the   crises   happened?   If   the   emerging   market   has   a   capital   flow   issue,   how   does   the   IMF’s  level  of  support  for  CFMs  vary?       Based   on   these   research   questions,   the   regression   models   are   built   on   the   following   equations  in  the  reduced  forms:   𝐼𝑀𝐹𝑑𝑖𝑎𝑔𝑛𝑜𝑠𝑖𝑠!" = 𝛼!" + 𝛽!" 𝐶𝑟𝑖𝑠𝑒𝑠! + 𝛾!" 𝑋!" + 𝜖!"     𝑀𝑒𝑛𝑡𝑖𝑜𝑛𝐶𝐹𝑀!" = 𝛼!" + 𝛽!" 𝐶𝑟𝑖𝑠𝑒𝑠! + 𝛾!" 𝑋!" + 𝜖!"     𝐿𝑒𝑣𝑒𝑙𝑜𝑓𝑠𝑢𝑝𝑝𝑜𝑟𝑡!" = 𝛼!" + 𝛽!" 𝐶𝑟𝑖𝑠𝑒𝑠! + 𝛾!" 𝑋!" + 𝜖!"   𝐼𝑀𝐹𝑑𝑖𝑎𝑔𝑛𝑜𝑠𝑖𝑠!"  is  coded  from  the  Article  IV  report  of  country  𝑖  at  year  𝑡  on  whether   there   is   a   mentioned   issue   of   capital   flows.   𝑀𝑒𝑛𝑡𝑖𝑜𝑛𝐶𝐹𝑀!"  is   coded   on   the   appearance   of   CFMs   after   there   is   an   initial   diagnosis   from   the   report.  𝐿𝑒𝑣𝑒𝑙𝑜𝑓𝑠𝑢𝑝𝑝𝑜𝑟𝑡!"  measures   quantitatively   the   attitude   of   IMF   towards   the   CFM   policies.  𝐶𝑟𝑖𝑠𝑒𝑠!  is   the   dummy   variable   which   takes   1   after   2008   and   0   otherwise.    𝑋!"  is   the   macroeconomic   fundamentals   that   measures   the   capital   vulnerability   of   the   emerging   markets.   Since   different   emerging   markets   have   different   economic   situations,   we   include   a   list   of   macro   fundamental   measures   as   the   fixed   effects   in   the   econometric   regression.   Our   selection   of   macro   variables   is   based   on   the   Economist   2013   Capital   Freeze   Index   (Economist,   2013).   We   make   a   modified   selection   of   the   variables   and   pick   first   three   key   elements:   current-­‐account   balance   as  %  of  GDP  (CAB),  short-­‐term  gross  external  debt  plus  external  debt  payments  as   %  of  foreign-­‐exchange  reserves  (EDP),  and  domestic  banking-­‐sector  credit  as  %  of   GDP  (DBC).     Current-­‐account  balance  is  defined  by  the  sum  of  the  value  of  imports  of  goods  and   services   plus   net   returns   on   investments   abroad,   minus   the   value   of   exports   of   goods  and  services.  When  a  country's  current  account  balance  is  positive  (surplus),   the   country   is   a   net   lender   to   the   rest   of   the   world.   When   a   country's   current   account   balance   is   negative   (deficit),   the   country   is   a   net   borrower,   making   the   domestic  economy  more  vulnerable  and  dependable  on  the  global  economy.   For   example,   South   Africa’s   current   account   deficit   is   high   relative   to   that   of   other   EMEs  and  is  financed  by  relatively  volatile  capital  inflows.  Foreign  direct  investment   has   typically   been   smaller   than   in   other   emerging   markets,   averaging   just   over   1   percent  of  GDP  in  the  past  ten  years  compared  to  around  3  percent  of  GDP  for  the   median   of   EMEs.   Instead,   South   Africa   has   been   more   reliant   on   portfolio   flows,   which   are   volatile   in   comparison   to   other   EMEs.   There   are,   nevertheless,   other     13   important  mitigating  factors.  External  debt  is  low  (26  percent  of  GDP  at  end-­‐2008),   over   40   percent   of   which   is   denominated   in   rand.   Banks,   corporations,   and   households   have   limited   foreign   currency   balance   sheet   exposure.   Capital   inflows   are  predominantly  in  the  form  of  equity,  and  hence  denominated  in  rand,  while  the   exchange   rate   floats.   Should   capital   outflows   reemerge,   foreign   investors   would   share   the   adjustment   burden—as   they   did   in   late   2008   when   the   stock   market   declined  and  the  rand  depreciated  sharply.   External   debt   is   that   part   of   the   total   debt   in   a   country   that   is   owed   to   creditors   outside   the   country.   The   more   debt   the   country   owes   the   outside   markets,   the   higher   risk   the   economy   bears   and   the   less   reliable   the   economy   is.   Domestic   credit   provided   by   the   financial   sector   includes   all   credit   to   various   sectors   on   a   gross   basis.  The  banking  sector  includes  monetary  authorities  and  deposit  money  banks,   as   well   as   other   banking   institutions   where   data   are   available.   Examples   of   other   banking   institutions   are   savings   and   mortgage   loan   institutions   and   building   and   loan  associations.  It’s  a  measure  of  the  health  of  the  banking  sector  of  the  economy   and  highly  related  with  the  vulnerability  of  the  capital  markets.   Here   we   first   use   current   account   balance,   domestic   banking-­‐sector   credit,   and   external  debt  payment  separately.  Second,  we  generate  a  composite  of  these  three   components  called  Capital  Vulnerability  Measure  (CVM)  following  the  rule  of  Capital   Freeze   Index   (CFI).   Then   use   CVM   as   one   single   independent   variable   in   our   regressions.  Third,  we  add  the  multiplication  term  of  the  control  variables  and  the   crises  dummy  in  the  regression.  This  aims  at  investigating  the  response  reasons  of   the  IMF  changes  after  the  crises.   The  regression  equation  is   𝑌!" = 𝛼!" + 𝛽!" 𝐶𝑟𝑖𝑠𝑒𝑠! + 𝛾!" 𝑋!" + 𝛿!" 𝑋!" ∙ 𝐶𝑟𝑖𝑠𝑒𝑠! + 𝜖!"   Where  𝑌!"  stands   for   capital   flow   diagnosis,   CFM   mention   or   support   for   capital   controls.   𝛿!"  measures   how   likely   IMF   is   to   respond   to   the   specific   economic   indicator  𝑋!"  with  𝑌!"  after  the  2008  crises.     The   reason   we   exclude   financial   openness   (Chin   and   Ito,   2008)   in   our   measure   CVM   is   that   there   exists   a   co-­‐linearity   between   CAB/EDP   and   financial   openness.   As   a   measure   of   financial   openness,   Chin-­‐Ito   index   is   a   summary   of   IMF’s   questions   on   countries  about  their  capital  accounts  which  includes  the  current  account  balance  of   the   country   and   other   related   variables.   Without   loss   of   generality,   we   also   adopt   the   Chinn-­‐Ito   capital   openness   index   separately   as   an   indicator   of   the   economy’s   capital  openness  to  see  the  impacts  on  IMF’s  institutional  view.   Last  but  not  least,  we  run  the  regression  for  different  regions  separately.  We  divide   the   emerging   markets   into   four   groups:   emerging   Asia,   emerging   Latin   America   and   Caribbean,   emerging   Europe   and   emerging   others.   Cross-­‐border   capital   flows   are   easier   and   more   frequent   inside   the   group   than   between   groups.   This   region   separation  can  be  regarded  as  the  regional  effects  in  capital  flow  liberalization.       14   2.2.2  Expected  Results   The   model   is   to   test   whether   the   crises   had   an   effect   on   IMF   advice   to   emerging   markets.  If  the  IMF  had  changed  its  advice  as  a  result  of  the  crisis  we  would  expect   to   observe   more   diagnoses   on   capital   flow   issues   after   2008   and   IMF’s   attitudes   towards   CFMs   are   more   supportive   than   before.   The   2008   global   financial   crises   have   a   tremendous   influence   on   the   emerging   markets,   especially   for   the   cross-­‐ border   capital   flows.   IMF’s   change   in   CFM   advice   will   result   in   significant   policy   changes  in  the  developing  countries.  Our  research  results  are  expected  to  give  solid   quantitative  evidence  on  the  IMF  policy  shifts  and  institutional  view  switches.   The   coefficient   of   Crises   measures   not   only   the   direction   of   the   changes   but   also   the   magnitude  of  the  effect.  We  test  on  whether  it’s  significantly  different  from  zero  and   interpret   as   a   shift   of   regime   after   the   crises.   A   comparison   between   whether   to   include   the   macro   fundamentals   is   helpful   in   understanding   these   changes   and   seeking   the   underlying   mechanism   which   causes   these   changes.   In   addition,   the   decision   on   the   choice   of   macroeconomic   variables   shed   light   on   how   different   channels  interrupt  with  each  other  and  which  part  affects  the  cross-­‐border  capital   flows  more  significantly.   Controlling   for   the   vulnerability   of   the   economy   improves   the   fitness   of   the   model   though   there   is   a   loss   on   the   number   of   observations.   The   more   vulnerable   the   emerging  market  is  (the  larger  the  capital  freeze  index),  we  would  expect  a  higher   chance   to   be   diagnosed   of   capital   flow   issues   initially,   the   less   likely   there   is   a   mention  of  CFM  due  to  the  instability  of  economy.  Otherwise,  the  coefficients  of  CVM   are  insignificantly  different  from  zero.  The  R-­‐square  indicates  the  goodness  to  fit  of   the   econometric   model.   Due   to   the   small   size   of   the   dataset   as   well   as   that   crises   might  not  be  the  major  reason  for  IMF’s  change  in  the  capital  flow  issue;  we  might   expect  a  small  R-­‐square.   The   regression   results   will   give   useful   suggestions   to   analyze   the   IMF’s   institutional   view   with   respect   to   the   capital   flow   liberalization   policy,   taking   the   economy   vulnerability   index   into   account.   Volatile   international   capital   flows   has   cross-­‐ border   financial   shocks   which   influenced   the   boom-­‐and-­‐bust   cycle   as   well   as   domestic   banking   credit.   It’s   necessary   to   keep   a   consistent   record   of   the   capital   flow   management   policies   and   learn   from   the   history   lessons   after   the   financial   crises.       15   3.    Results   3.1  Summary  Statistics We  present  summary  statistics  for  the  dataset  as  a  flow  chart  in  figure  4.  Figure  5   presents  summary  statistics  for  the  key  variables  used  in  the  analysis.  Figure  6  and   7   demonstrate   the   level   of   supports   through   a   breakdown   by   region   and   crises   event.     An   important   issue   is   the   representativeness   of   the   sample.   We   adopt   the   country   list  posted  by  IMF  authority  and  compare  our  coding  statistics  with  the  IMF  dataset3   between  1998  and  2000.  Our  dataset  mirrors  with  their  summary  table  by  90%.  The   results   show   that   our   sample   is   fairly   representative   in   terms   of   all   dimensions   (country,  region  and  coding  criteria).   3.1.1  Flow  Chart   Figure  4  Summary  statistics  of  coding  on  capital  flows   IMF  initial   diagnosis   • 78  obs  coded  0  as  capital  mlows  not  an  issue.   • 242  obs  coded  1  as  capital  mlows  mentioned.   Mention  of   CFMs   • Among  242  diagnosis,  144  obs  have  no  CFM  mention.   • 98  out  of  242  have  CFMs  mentioned  literally.   Level  of   Support   • Among  98  CFM  cases,  29  (30%)  are   coded  not  supportive.   • 58  (60%)  are  partially  supportive.   • 8  neutral  cases  and  3  fully   supportive.     Among   all   320   available   observation   of   IMF   initial   diagnosis   during   2001-­‐2013,   75.6%   of   Article   IV   reports   mentioned   capital   flows   as   an   issue   of   concern   for   the   country.   Besides,   CFMs   are   a   major   suggestion   as   40.5%   of   the   IMF   initial   diagnosis.   Though  capital  flow  management  is  considered  to  be  an  issue,  it  often  comes  along   with   macro   prudential   policies   and   sterilization/intervention   in   foreign   exchange   markets.   Therefore,   CFM   not   mentioned   for   a   diagnosis   is   not   equivalent   to   the   CFM’s   ineffectiveness   but   other   policies   alone   can   work   out   during   the   current   period.   With  a  mention  of  CFM,  61  out  of  98  cases  gain  supportive  advice  from  the  IMF  with   3   special   cases   of   fully   support.   There   are   only   8   neutral   cases   in   level   of   support,   which  indicates  IMF’s  CFM  policies  usually  come  with  a  degree  of  either  supportive   or   not   supportive.   It’s   of   interest   to   see   the   distribution   of   level   of   support   in   3  2012  IMF  new  institutional  view,  Chapter  4,  Table  2  and  Table  3.     16   different  years.  The  changes  in  the  structure  of  level  of  support  are  signals  of  IMF’s   policy   switches   therefore   help   understand   the   global   evolvement   of   capital   flow   liberalization.   Figure  5  Proportions  of  level  of  support  in  terms  of  total  mention  of  CFMs   Level  of  Supports  in  CFMs Mention  of  CFMs 14   12   10   8   6   4   2   0   200 200 200 200 200 200 200 200 200 201 201 201 201 1   2   3   4   5   6   7   8   9   0   1   2   3   Supportive   3   3   2   3   5   2   1   2   2   8   10   12   4   Neutral   0   1   1   1   0   0   0   0   1   1   1   0   0   Not  supportive   5   3   3   2   0   4   2   4   2   2   1   0   0     From  Figure  5,  we  can  summarize  that:   • • • • Mention   of   CFMs   (the   total   number   of   all   IMF   initial   diagnosis   with   CFM   mentioned)  became  more  frequent  since  2008.   The  level  of  support  in  capital  flow  liberalization  is  increasing  tremendously,   transferring   from   not   supportive   to   partially   supportive   or   even   fully   supportive.   Neutral   cases,   which   represent   no   attitude   toward   capital   flows   given   CFM   mentioned,  are  rare.  (2002-­‐2004  Chile,  2009  South  Africa,  2011  Venezuela).   The  level  of  support  is  more  relevant  after  2010  when  the  IMF  starts  its  New   Institutional  View.  Capital  controls  are  open  in  2005  and  2012.   The  number  of  countries  that  have  Article  IV  reports  before  and  after  crisis  has  been   balanced.   Before   the   crisis,   12.2%   of   the   countries   observed   gain   partial   support   for   capital   controls   from   the   IMF,   11.6%   of   the   countries   observed   gain   total   support,   while  after  the  crisis  the  proportions  are  22.3%  and  7%  respectively.                   17   Table  4  Correlation  between  different  policy  suggestions   Correlation   Fiscal   Policy   Exchange   Sterilization/   Trade   Macro-­‐ Rate   Intervention   Liberalization   prudential   Tighten  Fiscal  Policy   1.0000           Exchange  Rate  Flexibility   0.2395   1.0000         Sterilization/Intervention   0.0531   0.2788   1.0000       Trade  Liberalization   0.1018   -­‐0.0494   -­‐0.0774   1.0000     Tighten  Macroprudential   0.2480   0.2778   0.2375   0.0267   1.0000   Mention  of  CFMs   0.1131   0.1909   0.1971   0.0820   0.1981     Table   4   reports   correlation   coefficients   among   different   types   of   IMF   advice   on   capital   flows.     To   things   stand   out   that   our   relevant   to   this   study.     First,   Exchange   rate   flexibility   comes   along   with   sterilization/intervention   in   the   FX   market   and   tightening   prudential   regulations.     Second,   mention   of   CFMs   is   most   related   with   macro  prudential  regulation  policy.     3.1.2  Examples  from  Reports   The   major   concern   about   the   coding   process   is   the   difficulty   to   narrow   down   or   categorize   the   IMF’s   views   as   expressed   in   Article   IV   reports.   For   example,   the   report   might   make   a   reference   to   the   “liberalization   of   the   trade   and   exchange   system,”  which  may  or  may  not  include  capital  account  liberalization.  On  the  other   hand,   the   absence   of   an   explicit   reference   doesn’t   mean   that   the   IMF   never   expressed  a  view  during  the  policy  dialogue  meeting  process.    In  this  section  of  the   paper  we  note  illustrative  examples  of  diagnosis  and  support  that  were  coded  in  the   paper.       Capital  Flows  Mentioned  as  an  Issue:   • Guatemala  2006:  "Money  and  credit  expansion  has  been  rapid  as  a  result  of   high  liquidity  associated  with  strong  capital  inflows"(2#3);  “short-­‐term  goal   is  to  maintain  macroeconomic  stability”  (13#15).   • South   Africa   2009:   “The   global   financial   crisis   of  late   2008   sharply   changed   the   outlook   for   an   already   slowing   economy   and   posed   new   challenges   for   macroeconomic   policies.   Large   capital   outflows,   triggered   by   investor   withdrawal   from   emerging   market   assets   lowered   stock   prices   and   depreciated   the   rand.   South   Africa-­‐specific   factors,   such   as   the   high   current   18   account   deficit   and   policy   uncertainties   created   by   the   upcoming   national   elections   in   April   2009,   also   contributed   to   an   elevated   perception   of   risk”   (5#2.1).     • Colombia   2010:   “It   would   be   advisable   to   consider   possible   responses   to   a   surge  in  private  capital  inflows”  (11#16).   • Republic   of   Korea   2012:   “Reflecting   swings   in   global   risk   aversion,   capital   flows  into  Korea  have  been  highly  volatile.  Following  the  August–September   2011   spike   in   risk   aversion,   Korea   witnessed   a   surge   in   capital   outflows   of   around  US$15  billion  (including  outflows  from  European  banks)  in  the  third   quarter”  (5#7).       Thailand   2012:   “In   line   with   other   emerging   markets,   capital   flows   became   markedly   more   volatile   and   are   likely   to   remain   so   in   the   near   future”   (17#30).     Capital  Flows  Not  Mentioned  as  an  Issue:     • Guatemala   2012:   “Capital   inflows”   are   mentioned   but   “vulnerabilities   remain   limited”,   no   problems   are   foreseen   relating   to   capital   inflows   or   outflows   (2#2).     CAPITAL  FLOW  MEASURES   Mention  of  CFMs     • Thailand   2010:   “Relaxed   regulations   on   capital   outflows:   The   Thai   authorities   relaxed   regulations   for   domestic   residents   to   invest   in   foreign   securities   in   two   steps,   first   in   August   2009   and   then   in   February   2010,   to   encourage  capital  outflows  so  as  to  abate  the  upward  pressure  on  the  baht.   While   the   new   ceilings   are   far   from   binding,   Thai   investors   continue   to   accumulate  foreign  assets,  suggesting  that  the  measures  have  been  successful”   (BOX  2  p.23).     • South   Africa   2011:   “Authorities4'   Views:   Capital   Flow   Measures.   With   limited   scope  for  modifying  the  monetary  and  fiscal  settings  in  the  near  term  and  the   rand   on   the   strong   side   of   fundamentals,   there   is   arguably   a   case   for   using   either  an  unremunerated  reserve  requirement  or  a  small  tax  on  inflows  to  try   to  curtail  inflows  or  at  least  change  their  composition  (18#28.3);  As  for  CFMs,   the   authorities   agreed   with   staffs   that   their   effectiveness   is   questionable.   Moreover,   they   stressed   the   need   to   take   a   close   look   at   each   country’s   • 4  Authorities  vary  in  different  emerging  markets.  They  are  economists  and  researchers  from  local  central  banks  and   economic  research  institutes  as  well  as  professors  in  top  universities.     19   circumstances  in  deciding  on  the  adoption  of  CFMs.  In  South  Africa,  curtailing   inflows   while   the   country   is   also   relying   on   them   to   finance   domestic   consumption  and  investment  would  be  problematic  (20#29).      In  the  IMF  Article  IV  reports,  there  are  three   groups   of   authors:   the  authorities,  the   staffs,  and  the  directors.  Basically,  the  first  part  of  the  reports  is  drafted  by  the  staffs,   who   specialize   in   one   emerging   markets.   They   quoted   authorities’   views   from   economists   of   the   central   bank   of   the   country   and   gave   detailed   analysis   on   country’s   economy.   The   directors   gave   their   opinions   on   policy   and   wrote   the   second  part  of  the  reports.  We  obtain  the  level  of  supports  for  capital  controls  from   the   Directors’   view   under   the   assumption   that   the   staffs   and   directors   share   a   consistent  goal  in  advising  the  country’s  policy  decisions.  This  assumption  works  for   most   emerging   markets   cases   with   minor   fluctuations   due   to   systemic   difference   between  staffs  and  directors.       Level  of  Support  for  Capital  Controls   Examples  of  Not  Supportive:     • Malaysia   2001:   “Directors   also   welcomed   the   recent   removal   of   the   remaining   levy   on   profit   repatriation   of   portfolio   capital,   and   a   few   Directors   urged   the   authorities   not   to   resort   to   capital   controls   in   the   event   of   a   deteriorating  external  position  in  the  future."  (p.3,  PIN)     • Malaysia   2003:   “They   supported   the   authorities'   approach   to   gradually   liberalize  the  remaining  administrative  measures  on  capital  flows"  (p.5,  PIN)     • Malaysia   2011:   “There   is   also   no   evident   need   for   capital   controls   to   dampen   the  volatility  of  flows."  (7#14)     • Colombia  2007:  “[Capital]  controls  are  unlikely  to  be  effective  over  the  longer   term   and   are   at   odds   with   the   government’s   desire   to   deepen   Colombia’s   financial  markets  through  the  increased  participation  of  foreign  capital.  The   staff’s  preliminary  assessment  is  that  in  the  short  run,  the  controls  reduced   portfolio  inflows  and  borrowing”  (22#34)  &  in  regard  to  exchange  rate,  “Staff   welcomes   the   relaxation   of   capital   controls   undertaken   in   December,   and   recommends   that   the   authorities   consider   a   complete   phasing   out   of   the   controls  in  the  near  term,  given  their  limited  effectiveness”  (26#45)   • Colombia   2008:   “The   controls   were   also   associated   with   a   significant   increase  in  exchange  rate  volatility.  In  light  of  their  limited  effectiveness  and   adverse   effects   on   volatility   and   asset   market   development,   staff   saw   significant  drawbacks  to  the  controls”  (23#33)   20   Examples  of  Partially  supportive:     • Malaysia   2002:   “While   recognizing   that   capital   controls   have   played   a   role   in   helping  Malaysia  to  regain  financial  stability,  and  that  most  of  the  remaining   measures  are  mainly  capital  account  regulations  of  a  prudential  nature,  a  few   Directors5  encouraged  the  authorities  to  further  relax  them.”  (p.5,  PIN)     • South   Africa   2005:   “The   authorities   have   continued   with   their   gradual   approach   to   relaxing   capital   controls.   Staff   welcomed   the   easing   that   took   place  over  the  past  year,  including  the  removal  of  limits  on  outward  foreign   direct   investment.   The   main   remaining   restrictions   comprise   limits   on   overseas   investments   by   institutional   investors   and   the   prohibition   of   portfolio  investment  abroad  by  non-­‐financial  firms.  The  authorities  indicated   that   they   were   considering   moving   from   exchange   controls   to   a   system   of   prudential   regulations   for   institutional   investors.   Staff   supported   this   initiative  and  favored  a  further  easing  of  restrictions  on  non-­‐financial  firms.   In   its   view,   all   these   measures   would   increase   market   liquidity   and   allow   greater  risk  diversification,  and  could  reduce  currency  volatility  (15#29).     • Chile  2006:  “In  recent  years,  the  domestic  bond  market  has  grown  steadily,   but   it   remains   characterized   by   a   relatively   low   level   of   activity.   Staff   recommended  that  liquidity  be  improved  by  relaxing  some  of  the  investment   restrictions   on   the   private   pension   funds   and   reviewing   the   procedures   surrounding  the  taxation  of  foreign  investors.  Staff  also  suggested  removing   the  distortions  caused  by  the  stamp  tax  and  considering  introducing  a  system   of   specialists   in   public   debt   with   obligations   tailored   to   the   needs   of   the   market  (p.17#30);  improving  liquidity  in  the  capital  market  is  a  priority.  The   government   is   encouraged   to   develop   a   medium-­‐term   public   debt   strategy,   beyond   the   political   cycle,   and   decide   if   it   will  maintain   a   presence   in   issuing   bonds.   Regulations   are   needed   to   clarify   exemptions   from   capital   gains   tax   for   foreign   institutional   investors.   This   may   help   enhance   further   the   development   of   the   financial   sector   and   contribute   to   internationalizing   the   peso.”  (p.23#44)   • Thailand  2008:  “They  welcomed  the  removal  of  capital  controls,  providing  a   clear  signal  of  the  authorities'  intention  to  support  market-­‐friendly  policies.”   (EBA#2).   • Malaysia   2012:   “Dealing   with   capital   flows.   Malaysia   has   been   exposed   to   volatile   capital   flows,   and   this   is   likely   to   continue   in   the   near   term.   The   policy   response   so   far,   characterized   by   two-­‐way   exchange   rate   flexibility   while   smoothing   excessive   exchange   rate   fluctuations,   has   been   successful.     5  Directors  are  IMF  policy  chief  advisor  to  the  corresponding  country.     21   This   policy   should   continue   going   forward,   and   could   be   complemented   with   MPPs  if  needed  to  mitigate  potential  risks.”  (16#37)     • Colombia   2010:   “Staff   noted   that,   in   general,   the   effectiveness   of   controls   is   rather   limited   as   a   permanent   measure,   although   they   could   be   useful   to   manage   a   temporary   surge   in   capital   flows”   (12#16)   &   "If   the   response   were   also  to  include  some  type  of  capital  controls,  these  should  be  price-­‐based  and   applied  to  a  wide  range  of  transactions".  (23#37)   • Colombia   2011:   “Staff   noted   that   the   current   coverage   of   macro   prudential   measures  was  broad  and  adequate  and  that  CFMs  could  be  useful  as  part  of  a   transitory   response   to   an   unexpected   surge   in   inflows.   Staff   emphasized,   however,  that  a  tighter  fiscal  stance  would  have  to  be  the  main  element  of  the   policy   response   in   that   scenario.”   (18#19)   &   "Staff   welcomes   the   authorities’   plans   to   consider   other   policies   (including   strengthened   macro-­‐prudential   regulations  and  capital  flow  management  policies)  if  capital  inflows  are  seen   as   likely   to   endanger   financial   stability.   Staff   also   underscores   that   a   tightening  of  the  fiscal  stance  would  have  to  be  part  of  the  policy  response  to   a  possible  surge  in  capital  inflows."  (31#40)     Examples  of  Supportive:   • Malaysia   1999:   “Directors   broadly   agreed   that   the   regime   of   capital   controls-­‐ -­‐which  was  intended  by  the  authorities  to  be  temporary-­‐-­‐had  produced  more   positive  results  than  many  observers  had  initially  expected.  They  welcomed   the   pragmatic   and   flexible   way   in   which   Malaysia   had   implemented   and   adjusted  the  controls."  (pp.3-­‐4,  PIN)   • Turkey   1999:   “Directors   also   considered   it   appropriate   for   the   Central   Bank...to   counter   capital   inflows   by   allowing   interest   rates   to   fall.   In   this   context,  Directors  considered  the  proposed  reduction  in  the  ceiling  on  banks’   net   open   foreign   exchange   positions   to   be   a   positive   step   that   would   help   curb  capital  inflows  and  reduce  banking  sector  risk.“  (p.  4,  PIN)   • Chile  2000:  “They  also  supported  the  proposed  new  regulations  on  corporate   governance  and  the  recent  measures  to  liberalize  capital  flows,  citing  Chile  as   exemplifying  an  effective  and  well-­‐sequenced  approach  to  the  use  of  capital   controls  and  their  eventual  replacement  by  prudential  controls”  (EBA#5).   • South   Africa   2007:   “The   authorities   continue   to   relax   exchange   controls   gradually.   Remaining   controls   apply   to   capital   outflows   by   residents,   and   mainly  comprise  limits  on  overseas  investment  by  institutional  investors,  the   prohibition   of   portfolio   investment   abroad   by   corporates,   and   limits   on   offshore  investment  by  individuals.  Staff  supported  the  relaxation  of  controls,   as  it  allows  for  a  better  allocation  of  resources,  and,  by  deepening  the  foreign   22   exchange   market,   could   help   reduce   exchange   rate   volatility   over   the   medium   term.   Staff   favored   simplifying   the   administration   of   controls—for   instance,   replacing   prior   authorization   with   reporting   requirements   for   investments  within  the  allowed  limits—to  reduce  compliance  costs  (13#21).     • Republic   of   Korea   2012:   “Despite   this   progress,   Korea   is   subject   to   substantial   capital   flow   volatility   given   the   large   size   and   openness   of   its   capital   markets.   Even   though   the   banking   system   vulnerabilities   have   diminished  since  2008,  it  is  important  to  remain  vigilant  as  Korea’s  exposure   to   foreign   banks   on   the   funding   side   is   among   the   highest   in   Asia…   A   potential   deterioration   of   conditions   in   parent   banks’   jurisdictions   may   create   funding   difficulties   in   foreign   bank   branches,   which   will   need   to   be   monitored  closely”  (13#30).     3.1.3  Comparison  between  Regions   The   evolution   of   level   of   support   for   capital   controls   varies   by   region.   Country   heterogeneity   affects   our   model   and   causes   variation   in   the   results.   These   countries   differ  in  geographical  regions,  income  levels,  macroeconomic  fundamentals,  political   context,  and  the  size  of  the  country  and  so  on.  However,  we  notice  that  the  general   tendency   is   of   an   increase   in   the   level   of   support   over   years.   To   see   the   regional   heterogeneity,  we  carry  a  regional  analysis  of  the  change  in  level  of  support  over  the   years  in  depth.   Furthermore,  we  compare  the  mean  and  quarters  before  and  after  the  crisis  among   different  regions.  All  variables  are  significantly  different  before  and  after  the  crisis   with   the   exception   of   trade   liberalization.   Moreover,   in   general   all   variables   are   significantly   different   at   the   1%   level.   It   would   be   very   interesting   in   further   research  to  analyze  the  reasons  behind  the  changes  in  different  regions.   Figure  6  Comparison  of  box  distribution  of  level  of  support  before  and  after   crises     23   Figure   6   shows   the   general   distribution   of   level   of   support   among   different   emerging  markets.  The  box  boundary  represents  the  25  quantiles  and  75  quantiles.   The   lines   are   the   25%,   50%   and   75%   of   the   distribution   with   maximum   and   minimum   value   points.   For   emerging   Asia,   there   is   a   significant   spread   in   level   of   support  after  the  crises.  Emerging  Europe  has  a  significant  shift  from  not  supportive   to   fully   supportive.   Emerging   Latin   America   markets   are   comparatively   persistent   on   capital   flow   liberalization.   Emerging   others   has   mixed   issues.   All   emerging   markets  receive  more  volatile  level  of  support  on  capital  flows  from  IMF.   We   see   that   the   IMF   strengthens   support   for   partial   control   of   capital   account   significantly  after  the  crisis.  There  is  no  substantial  adjustment  on  the  level  for  not   supportive,   neutral   and   totally   supportive.   However,   without   controlling   for   other   vulnerability  indices,  it  is  not  safe  to  claim  that  the  financial  crisis  alters  the  IMF’s   support   level   for   capital   control.   It   is   highly   likely   that   the   IMF   changes   its   policy   recommendation   for   capital   controls   based   on   the   four   vulnerability   indices   we   introduced.     Figure  7  gives  a  frequency  graph  of  the  levels  of  support.  Here  “zero”    case  not  only   includes  a  neutral  attitude  towards  the  level  of  support  in  capital  controls,  but  also   accounts  or  the  cases  of  not  mentioning  CFMs  as  well  as  capital  flow  not  diagnosed   as  an  issue.  We  see  a  larger  proportion  of  capital  flow  unclearness  in  the  emerging   Latin   American   and   Caribbean.   On   the   contrary,   emerging   Asia   has   the   highest   frequencies  in  both  not  supportive  and  supportive  categories.     Figure  7  Statistics  of  Level  of  Support  by  Region  Including  All  Observations     24   4.    Regression  results   4.1  Baseline  model We   first   run   the   simple   regression   of   IMF   Initial   Diagnosis   on   the   Financial   Crisis   Dummy   variable.   Results   are   summarized   in   Table   5   Panel   1.   When   we   run   the   simple  regression  of  IMF  Initial  Diagnosis  on  dummy  variable,  we  obtain  a  positive   and  statistically  significant  coefficient  for  the  dummy  variable.  Without  control  for   other  variables,  there  is  a  statistically  significant  change  in  IMF  Initial  Diagnosis  due   to  the  2007-­‐2008  Financial  Crisis.   However,   the   conclusion   we   have   reached   by   just   running   the   previous   regression   doesn’t  control  for  any  other  variable  that  could  have  an  effect  on  the  significance  of   the  “Dummy”.  It  may  be  the  case  that  it  is  not  due  to  the  2008  financial  crisis  itself   but  to  other  elements  such  as  the  economic  fundamentals  of  the  emerging  markets.   Therefore,  to  further  understand  the  relationship  between  the  Crises  and  IMF  Initial   Diagnosis  we  run  further  regressions.   As   shown   in   Table   5,   even   if   we   control   for   macro   fundamentals,   which   measures   the  vulnerabilities  of  the  economy,  into  the  regression,  the  dummy  variable  always   has  a  statistically  significant  and  positive  impact  on  the  IMF  initial  diagnosis.     We  plot  the  baseline  regression  model  of  the  level  of  support  of  capital  controls  for   different   regions   in   Figure   8.   There   are   positive   trends   for   emerging   Asia   and   Europe.  The  grey  area  represents  the  90%  confidence  interval  for  the  slope.   Figure  8  Plot  and  fitted  OLS  for  the  level  of  support  by  region     25   4.2  Including  Capital  Vulnerability  Measures   Our   modification   of   the   baseline   model   includes   adding   current   account   balance,   domestic  banking  sector  credit  and  external  debt  payment  ratio  separately  as  well   as   capital   vulnerability   measures.   Capital   vulnerability   measure   is   generated   from   the  combination  of  three  variables  as  discussed  in  the  previous  sector.   First,   IMF’s   initial   diagnosis   changes   significantly   after   the   2007   financial   crises,   with   an   increase   of   around   20%   on   the   cross-­‐border   capital   flow   issues.   There   is   no   significant   impact   from   the   current   account   balance,   domestic   banking   credit   and   external  debt,  neither  separate  nor  simultaneous.     Second,   mention   of   CFMs   is   significantly   correlated   with   the   domestic   banking-­‐ sector   credit:   the   higher   the   credit   is,   the   more   vulnerable   the   domestic   banking   sector   is,   therefore   the   more   mention   of   CFMs   as   a   warning   advice.   The   Capital   Vulnerability  Measure  generated  from  the  three  variables  has  a  significant  positive   coefficient,   meaning   the   more   vulnerable   the   emerging   market   is,   the   higher   probability  (50%  increases)  of  mentioning  CFMs.     Third,   level   of   support   on   capital   flow   liberalization   changes   after   the   crises,   with   more   supportive   arguments   mentioned   in   the   Article   IV.   The   more   severe   the   current   account   deficit   (negative   balance),   the   less   supportive   IMF   is   on   capital   flow   liberalization.  What’s  more,  for  the  domestic  credit  from  the  banking  sector,  a  high   value  of  domestic  credit  implies  a  high  development  in  capital  market  together  with   a   high   risk   of   capital   vulnerability.   In   2013,   the   country   with   the   highest   domestic   credit   in   the   world   is   Japan   with   a   value   of   341.69   while   the   lowest   value   in   the   world  is  Libya  -­‐65.93.  We  find  the  positive  relationship  between  domestic  credit  and   mention   of   CFMs   because   a   boom   in   banking   credit   makes   IMF   realize   that   the   country   needs   to   restrict   inflows.   IMF   is   more   cautious   to   the   capital   flows   in   the   emerging  markets,  valuing  domestic  credit  as  a  signal  of  economy’s  vulnerability.     Taking  one  step  further,  we  generate  Capital  Vulnerability  Measures  from  previous   results  and  regard  it  as  a  general  control  variable  of  the  economy’s  vulnerability.  In   the  final  regression,  there  is  a  positive  correlation  between  the  vulnerability  of  the   emerging   market   and   the   supportive   attitude   of   IMF.   The   more   vulnerable   the   current   economy   is,   the   more   changes   in   IMF’s   level   of   support   on   capital   flows   management  measures.  These  changes  can  be  from  not  supportive  to  neutral,  from   neutral  to  partially  supportive,  or  from  partially  supportive  to  fully  supportive.   The  R—squares  of  our  regression  models  are  low  because  our  sample  is  relatively   small  and  has  a  short  time  horizon  of  12  years.  By  adding  more  control  variables,  we   see   an   increase   in   the   model’s   goodness   to   fit.   There   are   two   kinds   of   effects   by   including   more   control   variables:   on   one   hand,   it   lowers   the   degree   of   freedom   of   the  model;  on  the  other  hand,  it  avoids  the  model  misspecification  and  improves  the   model   explanation.   More   observations   would   be   helpful   to   better   explain   the   effects   of  financial  crises  on  IMF’s  view  of  capital  flows.     26   4.3  Including  Financial  Openness  alone   Table   6   shows   the   results   of   adding   financial   openness   alone   in   the   regression   model.   When   using   Chin-­‐Ito   Capital   Openness   as   the   control   variable   of   the   underlying   economic   fundamentals,   there   is   a   significant   negative   effect   on   both   IMF’s   initial   diagnosis   and   mention   of   CFMs.   The   coefficient   can   be   interpreted   as   the   influence   of   capital   openness:   the   higher   level   of   openness   the   current   emerging   market  is,  the  less  probability  it’s  diagnosed  by  IMF  with  a  capital  flow  issue  and  the   less  mention  of  CFMs  in  the  IMF  Article  IV  report  of  the  same  year.   4.4  Adding  intersections   Beyond  our  baseline  model,  we  generate  five  intersection  variables  by  multiplying   the  controls  with  the  crises  dummy.  There  are  dummycurrentbalance,   dummydomesticcredit,  dummyexternaldebt,  dummyfinancialopenness  and   CCompositeindex.  CCompositeindex  is  based  on  Capital  Vulnerability  Measures  we   created,  which  is  the  equally  weighted  composite  index  of  current  account  balance,   external  debt  and  domestic  banking-­‐sector  credit.  By  adding  these   𝑐𝑜𝑛𝑡𝑟𝑜𝑙  𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒×𝑐𝑟𝑖𝑠𝑒𝑠  𝑑𝑢𝑚𝑚𝑦   Variables,  we  expect  IMF  is  more  likely  to  respond  to  specific  economic  indicators   with  a  capital  flow  diagnosis,  CFM  mention  as  well  as  support  for  capital  controls   after  the  2008  crises.  There  should  be  a  positive  significant  coefficient  for  the   intersection  variables.   The  main  purpose  of  adding  the  intersection  parts  is  to  find  out  how  IMF  makes   changes  and  adjustments  in  capital  flow  regulations  after  the  2008  financial  crises.   As  we  have  already  shown  in  our  quantitative  analysis,  IMF  does  change  its  level  of   support  to  CFMs  after  the  crises.  But  what  are  the  key  factors  IMF’s  changes  in  its   initial  diagnosis,  mention  of  CFMs  and  level  of  supports  are  based  on?  Which  macro   indicators  influence  IMF’s  change  most  after  the  crises?  Are  the  changes  in  different   decisions  affects  by  the  same  factor  significantly  and  consistently?       The  results  are  shown  in  table  7.  There  are  some  interesting  results  we  found   through  tens  of  regressions  we  run.  Most  regressions  turn  out  to  be  not  significant.   However,  there  are  improvements  in  the  model’s  goodness  to  fit  and  some  of  the   results  come  out  to  be  exactly  the  same  as  our  expectation.     We  found  significant  positive  response  in  the  financial  openness  index  to  IMF’s   initial  diagnosis  after  2008  and  negative  for  the  level  of  support.  There  are  more   IMF  initial  diagnoses  after  the  crises,  which,  with  a  higher  probability,  is  due  to  the   financial  openness  of  the  emerging  markets.  IMF’s  judgment  on  capital  flows  issues   is  mostly  based  on  countries’  financial  openness.  However,  the  mention  of  CFMs  is   more  related  to  the  current  account  balance.  Both  the  external  debt  and  financial   openness  have  negative  influence  on  IMF’s  levels  of  support  to  CFMs,  which  means     27   IMF  considers  external  debt  and  financial  openness  when  giving  opinions  on  the   policy  support.   For  mention  of  CFMs,  the  current  account  balance  matters  more  for  such  changes.   External  debt  has  a  significantly  negative  impact  on  the  level  of  support  changes   after  the  crises.  The  measure  of  capital  vulnerability  matters  more  for  the  IMF  initial   diagnosis  after  the  crises   In  fact,  our  capital  vulnerability  measure  turns  out  to  offer  a  relatively  consistent   result,  the  same  as  what  we  expected  in  the  beginning.  The  rest  of  the  regression   results  are  not  significant  from  zero  and  hard  to  tell  the  sign  of  the  coefficients.   4.5  Findings  and  Robustness   We  can  conclude  from  the  regression  results  that  the  financial  crisis  has  a  significant   influence   on   the   IMF’s   decision   about   level   of   support   for   capital   control   after   controlling   for   the   vulnerability   measures   individually.   Domestic   credit   in   the   banking  sector  and  the  CVM  index  both  alter  IMF’s  support  level  for  capital  control   significantly  in  the  expected  direction.  The  positive  coefficient  indicates  that  as  the   economy   becomes   more   vulnerable,   the   level   of   support   for   capital   control   increases.       The   result   is   especially   meaningful   for   emerging   market   economies   in   Asia,   which   are  mostly  under  development  with  credit  issues  in  the  banking  sector.  Capital  flows   can   be   associated   with   the   domestic   intermediary   sectors   such   as   banking.   In   particular,   positive   net   flows   can   be   used   to   finance   current   account   deficits.   In   August  2013,  India  announced  a  new  capital  control  to  stop  the  cash  flowing  out  of   the   country   and   to   stem   the   decline   of   rupee.   Since   our   regression   results   show   that   IMF   has   altered   its   policy   recommendation   after   the   financial   crisis,   India’s   imposition  of  the  capital  control  should  have  gained  support  from  IMF.     The   coefficient   on   capital   vulnerability   measure   CVM   is   positive   and   significant,   indicating  that  the  openness  of  a  country’s  capital  account  does  have  an  impact  on   IMF’s  altitude  about  capital  control.  However,  considering  the  fact  that  the  Chinn-­‐Ito   index  is  just  one  method  to  describe  certain  facets  of  a  country’s  capital  account,  we   cannot   say   that   in   reality   the   openness   of   a   country’s   capital   account   has   no   influence  on  IMF’s  policy  recommendation.     The   coefficient   on   vulnerability   index   “Current   account   balance”   is   negative   but   insignificant,   which   implies   that   we   cannot   reject   the   hypothesis   that   current   account   balance   does   not   alter   the   IMF’s   attitude   about   capital   controls.   The   coefficient   on   vulnerability   index   “External   Debt   Payment”   is   negative   but   insignificant,   which   implies   that   we   cannot   reject   the   hypothesis   that   the   ability   of   a   country  to  repay  its  debt  does  not  alter  the  IMF’s  attitude  about  capital  controls.     The  interesting  result  is  that  the  coefficient  on  the  domestic  banking-­‐sector  credit  is   negative   and   significant,   which   means   that   the   domestic   credit   of   a   country   in   its     28   banking  sector  alters  the  IMF’s  attitude  about  capital  controls  after  the  crisis.  After   the  crisis,  as  the  domestic  banking  credit  evaluated  by  World  Bank  strengthens,  the   IMF  would  be  more  likely  to  increase  its  level  of  support  for  capital  controls.   Our  results  can  be  summarized  as  follows.     • • • The  financial  crises  had  a  significant  impact  on  IMF  diagnosis  of  whether   capital  flows  are  a  source  of  vulnerability  in  emerging  markets.    This  finding   is  irrespective  of  the  level  actual  capital  flow  vulnerability  in  specific   economies  however  and  thus  signals  an  ideational  change.  22.9%  of  more   capital  flow  diagnosis  appears  after  the  crises.     The  IMF  is  more  apt  to  outrightly  discuss  CFMs  after  the  crises,  especially   when  domestic  banking  sector  credit  appears  to  be  concerning.  By   controlling  the  vulnerability  of  the  economy,  the  effect  of  crises  becomes   insignificant.  Our  capital  vulnerability  index  has  a  significant  prediction  of   IMF  mention  of  CFMs.     The  IMF  is  more  apt  to  support  the  use  of  CFMs  after  the  crisis,  however  the   level  of  support  changes  after  the  crises  becomes  less  significant  when   adding  vulnerability  controls.  The  more  vulnerable  the  emerging  market  is,   the  more  level  of  support  on  capital  flow  management  measures  are  imposed.     Thus  we  can  conclude  that  the  financial  crises  changed  the  pattern  of  IMF  attitudes   on  capital  flow  policies.  The  macro  fundamentals  of  the  economy  also  influences   IMF  view  on  capital  controls,  especially  domestic  banking-­‐sector  credit.  The  capital   vulnerability  index  significantly  affects  the  CFMs  suggestions  and  IMF  level  of   support.    Indeed,  the  link  between  the  crisis  and  IMF  support  begins  to  weaken  the   more  that  macroeconomic  control  variables  that  are  introduced  into  the  model.   The   results   are   robust   and   consistent   under   several   modifications   from   the   baseline   model.   Adding   intersections   and   more   controls   improves   the   model’s   goodness   to   fit.   The   main   results   don’t   change   and   the   corresponding   coefficients   remain   significantly  positive.  The  joint  endogenous  relationship  between  CFM  mention  and   levels   of   support   enhances   the   IMF   institutional   view   changes   after   the   Global   Financial  Crises.  Our  results  shed  light  on  the  further  research  on  what  factors  the   good   indicators   of   IMF   changes   are   and   seeking   the   underlying   economic   reasons   for  emerging  market  growth.             29   5.  Summary  and  Conclusions   This  paper  sought  to  add  to  the  existing  literature  on  the  IMF  and  the  capital   account  by  econometrically  testing  whether  the  financial  crisis  was  linked  to  a   change  in  IMF  policy  advice  on  these  matters.    The  IMF  underwent  a  significant  re-­‐ evaluation  of  its  policy  on  capital  account  liberalization  and  the  role  of  capital   controls  in  the  wake  of  the  global  financial  crisis.    Previous  work  has  shown  that  this   shift  in  thinking  at  the  IMF,  albeit  an  incremental  one,  was  due  to  a  number  of   factors  (Chwieroth,  2012;  Gallagher,  2014).    This  paper  adds  to  that  literature  and   finds  that  not  only  has  the  IMF  changed  what  its  view  on  capital  flows,  there  is  also   evidence  that  the  IMF  has  also  changed  its  actual  behavior  on  these  matters  as  a   result  of  the  crisis.       This  paper  is  not  the  last  word  on  these  matters.    Our  database  will  need  to  be   updated  on  an  annual  or  semi-­‐annual  basis.    In  so  doing  it  will  be  interesting  to   examine  the  extent  to  which  the  IMF  view  remains  a  significant  component  of  IMF   advice  as  the  ‘salience’  of  the  crisis  wanes  in  future  years.    Pagliari  (2013)  has   shown  that  policy-­‐makers  tend  to  be  most  attune  to  regulations  during  and  in  the   immediate  aftermath  of  crises  but  that  such  attention  decreases  as  the  public  and   policy-­‐makers  move  on  to  other  concerns.    What  is  more,  a  significant  amount  of   additional  coding  could  be  done  to  expand  the  set  of  independent  variables  in  an   attempt  to  explain  IMF  policy  on  capital  flows.    A  number  of  authors  have  coded  the   training  of  IMF  employees  and  economists  to  examine  how  the  ideas  and  beliefs  of   staff  are  reflected  in  policy  (Chwieroth,  2009;  Ban,  2013).    Adding  analyses  such  as   those  may  shed  further  insight  into  these  dynamics  as  well.   30